[tune] ResourceChangingScheduler dynamic resource allocation during tuning
#16787
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@richardliaw @krfricke Could you take a quick look over and let me know if I am on the right track? Thanks! |
richardliaw
reviewed
Jul 1, 2021
| self, trial_runner: "trial_runner.TrialRunner", trial: Trial, | ||
| resources: Union[Dict, Callable, PlacementGroupFactory]): | ||
| print(f"setting trial {trial} resource to {resources}") | ||
| trial_runner.update_trial_resources(trial, resources) |
There was a problem hiding this comment.
Can we avoid doing this in the scheduler and instead rely on the event loop to update the resources?
Specifically, something like:
action = on_result(runner)
# pause trial and update resources
...
# when cluster has available resources, they should start trial from checkpoint.
trial = choose_trial_to_run(...)
There was a problem hiding this comment.
@richardliaw changed - let me know if it's better now
Co-authored-by: Kai Fricke <[email protected]>
Co-authored-by: Kai Fricke <[email protected]>
|
@richardliaw This is almost done implementation-wise. Needs docs & tests. Updated the description. |
Yard1
changed the title
ResourceChangingScheduler dynamic resource allocation during tuning
|
@richardliaw Docs have been added and code cleaned up. Ready for review! 🚀 |
richardliaw
reviewed
Jul 6, 2021
Comment on lines
139
to
142
| total_available_cpus = ( | ||
| trial_runner.trial_executor._avail_resources.cpu) | ||
| total_available_gpus = ( | ||
| trial_runner.trial_executor._avail_resources.gpu) |
There was a problem hiding this comment.
hmm, maybe we want to expose this a bit better
richardliaw
reviewed
Jul 6, 2021
krfricke
reviewed
Jul 9, 2021
krfricke
reviewed
Jul 12, 2021
|
Looks good, ping when tests pass? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Why are these changes needed?
This PR adds a scheduler and API needed for dynamic resource allocation in Tune.
Includes #16844
One idea for improvement is prioritisation of placement groups. This would need to be implemented on the core's side. Regardless, the current setup seems to be working well.
Known issues:
Related issue number
Checks
scripts/format.shto lint the changes in this PR.